Agent consistently times out with local Ollama (mistral:7b-32k) on Windows — config changes don't help

Agent consistently times out with local Ollama (mistral:7b-32k) on Windows — config changes don't help

Hi everyone, I'm stuck with a persistent agent timeout issue and would really appreciate some pointers.

Environment

OS: Windows 10 (64-bit), i9-14900K, 64 GB DDR5

OpenClaw version: 2026.4.21 (installed via official PowerShell script)

Ollama version: latest, with mistral:7b-32k and mistral:7b pulled

Gateway: running as local Scheduled Task, loopback

What works

ollama run mistral:7b-32k replies quickly from the terminal (no issue with the model itself).

Gateway starts cleanly and shows [gateway] agent model: ollama/mistral:7b-32k.

openclaw skills install and clawhub commands work fine.

What fails

TUI (openclaw tui) shows Bootstrap pending and never completes, even after creating a valid BOOTSTRAP.md.

openclaw infer model run with the same model always times out after 2 minutes:

text

复制

下载

[agent/embedded] embedded run failover decision: ... decision=surface_error reason=timeout from=ollama/mistral:7b-32k

openclaw agent --message behaves identically.

What I have already tried (no effect)

Set agents.defaults.llm.idleTimeoutSeconds to 0 (and to 600) in openclaw.json.

Deleted sessions.json and the whole agents/main/sessions folder to force a fresh session.

Switched the primary model between mistral:7b and mistral:7b-32k.

Reinstalled skills, ran openclaw setup and openclaw doctor.

Tried the cron approach — the job was created but never produced output.

Question Is this a known limitation of the embedded agent engine with local Ollama models on Windows? Are there any hidden timeouts or workarounds I might have missed?

Thanks in advance for any help!Agent consistently times out with local Ollama (mistral:7b-32k) on Windows — config changes don't help

Hi everyone, I'm stuck with a persistent agent timeout issue and would really appreciate some pointers.

Environment

OS: Windows 10 (64-bit), i9-14900K, 64 GB DDR5

OpenClaw version: 2026.4.21 (installed via official PowerShell script)

Ollama version: latest, with mistral:7b-32k and mistral:7b pulled

Gateway: running as local Scheduled Task, loopback

What works

ollama run mistral:7b-32k replies quickly from the terminal (no issue with the model itself).

Gateway starts cleanly and shows [gateway] agent model: ollama/mistral:7b-32k.

openclaw skills install and clawhub commands work fine.

What fails

TUI (openclaw tui) shows Bootstrap pending and never completes, even after creating a valid BOOTSTRAP.md.

openclaw infer model run with the same model always times out after 2 minutes:

text

复制

下载

[agent/embedded] embedded run failover decision: ... decision=surface_error reason=timeout from=ollama/mistral:7b-32k

openclaw agent --message behaves identically.

What I have already tried (no effect)

Set agents.defaults.llm.idleTimeoutSeconds to 0 (and to 600) in openclaw.json.

Deleted sessions.json and the whole agents/main/sessions folder to force a fresh session.

Switched the primary model between mistral:7b and mistral:7b-32k.

Reinstalled skills, ran openclaw setup and openclaw doctor.

Tried the cron approach — the job was created but never produced output.

Question
Is this a known limitation of the embedded agent engine with local Ollama models on Windows? Are there any hidden timeouts or workarounds I might have missed?

Thanks in advance for any help!